Multiple gateway servers for one StoreOnce CoFC repository?

[Vejete]

Is it possible to configure five gateways to talk to one Catalyst store via Fiber Channel? We have five proxy servers for reading snapshot data from primary storage. I would like each of the five proxies to also be gateway servers to StoreOnce, but I don't really want to create five repositories to make that happen. For a single StoreOnce CoFC repository, it's not obvious to me how to establish access from multiple gateways to one Catalyst store. All advice appreciated.

[Andreas Neufert]

I will ask our HPE SAs to comment about this.

For performance reasons we always recommend to use Ethernet with StoreOnce.

[Vejete]

Thank you for your response Andreas. My cu has 1 GbE and 16 Gb FC, so I suspect he will fancy the faster SAN fabric rather than slower Ethernet.

Upon further investigation yesterday, I noted that the automatic gateway option is dimmed the moment you select the Fiber Channel option. I can see no technical reason for this exclusivity since the cu has five media servers all zoned to and successfully communicating with the StoreOnce. I suspect Veeam's behavior in this area was coded to protect against situations where one or more gateway servers may not have fabric access to the StoreOnce via the SAN.

The better solution would seem to be allowing automatic gateway selection in combination with offering the Catalyst transport options of Ethernet only, Fiber Channel only, or FC preferred with fallback to Ethernet. That is what other ISV backup apps offer.

You've also piqued my curiosity with regard to stating Ethernet is preferred for performance reasons. Why is that? If you put Ethernet and FC head-to-head for mass storage access, FC will win every time. FC has a much larger frame size and measurably lower protocol overhead compared to Ethernet. I'll take 16 Gb or 32 Gb FC mass storage access all day long instead of 10 GbE.

[foggy]

Automatic gateway selection is not available since we do not know which servers have FC access to repository unless user tells us explicitly. So currently you have to create multiple repositories if you want to make use of several gateway servers.

[Vejete]

Thanks, Foggy. That's the conclusion we've come to. Let's hope that changes in VBR v10. The StoreOnce Catalyst Client API contains a call that lets you query info including the COFC identifier via IP. That's why Micro Focus Data Protector lets you provide an appliance IP then specify whether you want IP only , FC only, or FC preferred with IP fallback. Then it doesn't matter who has fabric access and who does not. If fabric access is available on a gateway, it gets used. Otherwise, IP handles the datacomm. It's not rocket science. Veeam simply needs to implement the API call.

Then again, we're all still waiting (and waiting and waiting) for Veeam to utilize the Catalyst Copy API functionality so that backup copy moves deltas directly from source appliance to destination appliance in a low-bandwidth fashion rather than needlessly rehydrating data, dragging it across the LAN, and jamming it through one or more hosts as it does now.

[foggy]

It's not about the appliance, but the gateway. Veeam B&R knows appliance COFC, but it cannot check if the gateway server has FC access to the storage.

[dbewernick]

Just to clarify: Is it the same for DataDomain with multiple gateways via FC to a single pool?

[foggy]

Yes, it's the same for Data Domain.